With the progress of science and technology and the increasingly fierce competition in the market, the phenomenon of homogenization of jewelry products is becoming more and more obvious, and the user's demand for jewelry design presents personalized and diversified characteristics. This paper utilizes continuous variable binary to encode the genes in jewelry design and establishes the specific definition domain for each module in jewelry design. Then the fitness function is combined with the target creative set and representative sample set of jewelry design to calculate the consensus and satisfaction degrees of the audience. Finally, the construction of the jewelry design system is realized using artificial intelligence through selection, crossover, mutation, and other operations in the metagenetic algorithm. The algorithm proposed in this paper can achieve a lower number of evaluations and a shorter running time in a jewelry design system, which can alleviate user fatigue more effectively. When the population is generated over 20 generations, the average adaptation value of the algorithm in this paper can reach 8.13, which is much higher than the comparison algorithm. In the subjective evaluation experiments of the jewelry products obtained by the design, it is found that the audience thinks that the jewelry products show higher emotional potency and innovative value, which indicates that the system is able to give higher creativity and value to the jewelry products. This paper provides new foundations and insights for the field of jewelry design, as well as scientific ideas for jewelry designers and brands.